G Protein-Coupled Receptor-Ligand Pose and Functional Class Prediction.

Area of Science:

• Pharmacology
• Computational Chemistry
• Biochemistry

Background:

• G protein-coupled receptors (GPCRs) are vital transmembrane proteins and frequent drug targets.
• Virtual screening (VS) is extensively used in drug discovery programs targeting GPCRs.
• Enhancing the accuracy of predicting molecule binding and function for GPCRs can accelerate drug discovery.

Purpose of the Study:

• To evaluate the advantage of ligand interaction fingerprints over automated methods for binding site selection in docking.
• To determine if ligand interaction fingerprints can predict the functional status (agonist, antagonist, inverse agonist) of candidate molecules using a random forest classifier.

Main Methods:

• Classical docking simulations were performed.
• Ligand interaction fingerprints were assessed for their utility in binding site selection and pose sampling.
• A random forest classifier was trained and tested using ligand-receptor complex data to predict ligand function.
• The classifier's performance was evaluated on an external test set of GPR31 and TAAR2 ligands.

Main Results:

• Ligand interaction fingerprints provided modest advantages in sampling accurate poses but no substantial benefit in top-ranked poses after scoring.
• The random forest classifier effectively predicted ligand function, classifying agonists, antagonists, and inverse agonists as active.
• The binary classifier achieved an 82.6% hit rate for actual actives within the predicted active set on an external test set.

Conclusions:

• Ligand interaction fingerprints are not essential for generating high-quality ligand-receptor complexes for GPCR drug discovery.
• Random forest classification using ligand interaction fingerprints is a highly effective method for predicting GPCR ligand functional status.
• This predictive capability can significantly accelerate the identification of novel GPCR-targeting drug candidates.